1. Field
The present disclosure relates to a technique for calibrating an optical receiver.
2. Related Art
In a typical optical receiver, an optical signal is received by a photodiode, which generates current that is amplified by a current-to-voltage converter. This photodiode generates the current when the optical signal is present.
Typically, the optical signal is modulated by a transmitter with a finite extinction ratio, which results in an optical signal that is not fully turned ‘on’ or ‘off.’ Instead there is a finite ratio of the optical-signal levels, with an average DC level between them. As a consequence, the current has two components, an average current (Iavg) and a delta- or signal current (Isig) superimposed on Iavg (usually, Isig is a fraction of Iavg). For binary data, the transmitter modulates an optical source so that the photodiode outputs either Iavg−Isig (which corresponds to a logical ‘0’) or Iavg+Isig (which corresponds to a logical ‘1’).
Moreover, the optical path also usually introduces attenuation to the optical signal levels that are not known a priori to the time of assembly, and which can vary during normal operation of an optical link. As a consequence, the average DC level of the optical signal (and thus, Iavg) can vary over a wide range of possible values.
A current-to-voltage converter in an optical receiver typically includes one or more gain stages followed by a decision circuit (such as a digital slicer), which converts an analog current into digital voltage levels. Because the gain stage(s) usually has a narrow window of operation, the variation in the average DC level can pull the current-to-voltage converter away from its optimal DC-gain point.
Consequently, a crossover voltage (Vc) of the output-voltage swing and the biasing of the current-to-voltage converter are typically adjusted to ensure that the output voltages from the optical receiver correctly correspond to a logical ‘1’ (when the current-to-voltage converter outputs a high voltage, Vhi) and a logical ‘0’ (when the current-to-voltage converter outputs a low voltage, Vlow). In existing optical receivers, adaptive adjustment of a reference voltage (Vref) to which the output voltage is compared (which is sometimes referred to as ‘centering’) and the biasing of the current-to-voltage converter (which is sometimes referred to as ‘calibration’) are performed in a variety of ways. For example, an RC (low-pass) filter can be used to obtain Vref from the output voltage. However, because the RC filter passes low frequency information, in these techniques the optical signal is typically DC balanced. Furthermore, the RC-filter bandwidth typically needs to be high enough to track changes in the DC level of the output voltage, while being low enough to provide a stable value of Vref between bits.
In some existing optical receivers, DC-balanced codes (such as 8/10 codes) are used to significantly reduce the RC-filter bandwidth at the cost of added latency. Alternatively, if the environmental variations are slow, periodic calibration can be performed, and Vref can be stored on a capacitor. However, this technique may be difficult to implement due to leakage current from the capacitor and the inability to scale an on-chip capacitor.
In other existing optical receivers, biasing and Vref are obtained from the output voltage using an RC-feedback circuit that includes an RC filter. However, it may be difficult to stabilize the feedback loop while ensuring that it is fast enough to track changes. Even if the feedback loop is stable, relative to other approaches these optical receivers typically include additional switches and capacitors to prevent loading of the current-to-voltage converter output and to store the bias voltage, which increase the cost, power consumption and complexity of the optical receivers.
Hence, what is needed are an optical receiver and a calibration technique without the above-described problems.